Applying tvdi based on remote sensing data to evaluate the drought in Cu Chi district

41 Vietnam Journal of Hydrometeorology, ISSN 2525-2208, 2020 (04): 41-52 Tran Thi Thanh Dung1, Duong Thi Thuy Nga1 ABSTRACT Drought is a constant threat to Vietnam which causes great damage to the economy as well as forest ecosystems. Due to the increasingly com- plex drought-related impacts, remote sensing technology with outstanding advantages com- pared to traditional research methods has been applied effectively in research, monitoring, and coping with drought. Normalized Difference Vegetation Index (NDVI) and Land Surface Tem- perature (LST) were calculated from Landsat im- agery. The Temperature Vegetation Dryness Index (TVDI) with the combination of LST and NDVI index, was used as an indicator for drought risk assessment in Cu Chi District in 2005, 2010, 2015, and 2020. The results show a significant increase in dry areas between 2005- 2010 and 2015-2020. On the other hand, the re- sults of the TVDI index and mapping drought of Cu Chi district on February 13, 2005, February 11, 2010, January 24, 2015 and February 23, 2020 are a basis for risk assessment and drought monitoring. Keywords: TVDI, Landsat 8, Drought, Cu Chi District 1. Introduction Drought is a severe natural disaster around the world, which is a complex, and slow-onset phenomenon that affects more people than any other natural hazard and results in serious eco- nomic, social, and environmental impacts (Belal et al., 2012). Drought affects both developed and developing countries, but in different ways (Wardlow et al., 2012). In Vietnam, droughts occur across the country at different rates and times, causing enormous economic and social losses, especially for water sources and agricul- tural production. So that monitoring drought is very important. On the other hand, droughts often occur on a large-scale, so the monitoring and research by the traditional approaches for drought monitoring that uses ground-based data are laborious, difficult, and time-consuming (Prasad et al., 2007). In addition to recent ad- vancements in the field of earth observation through different satellite based remote sensing sensors have provided researches continuous monitoring of soil moisture at a global scale, which can support drought assessment/monitor- ing. Remote sensing can be applied on a large Research Paper APPLYING TVDI BASED ON REMOTE SENSING DATA TO EVALU- ATE THE DROUGHT IN CU CHI DISTRICT ARTICLE HISTORY Received: March 20, 2020 Accepted: April 22, 2020 Publish on: April 25, 2020 TRAN THI THANH DUNG Corresponding author: trttdung@hcmus.edu.vn; dttnga@hcmus.edu.vn 1Ho Chi Minh City University of Science, Vietnam National University Ho Chi Minh City                            DOI:10.36335/VNJHM.2020(4).41-52 42 scale, all weather monitoring and multi-band working which are suitable for real-time moni- toring on a large-scale. In recent years, with the development of multi-temporal and multi-spec- tral remote sensing technologies, the large amount of observational data has been achieved, which made it possible for real-time drought monitoring (Huang et al., 2011). Currently, methods of remote sensing for drought monitor- ing include thermal inertia, microwave remote sensing and the vegetation indices, etc. The Satellite-derived drought indicators calculated from vegetation index and other surface param- eters other have been widely used to study droughts such as the Vegetation Condition Index (VCI), and Temperature Condition Index (TCI), TVDI. Kogan (1990, 1995) monitored drought by used the Vegetation Condition Index (VCI) and obtained good results from NOAA polar-or- biting satellite data. Moran et al. (1994) sug- gested Water Deficit Index (WDI) by extending Crop Water Stress Index (CWSI) to partly veg- etation cover conditions. The Vegetation Tem- perature Condition Index (VTCI) is a near real-time approach of drought monitoring that is related to the NDVI and the LST changes devel- oped by Wang et al. (2001). Sandholt et al. (2002) proposed a simplified soil surface dryness index based on an empirical parameter of the re- lationship between Ts and NDVI to detect the drought levels based on a large amount of data remote sensing called TVDI. Wang et al. (2004) evaluated the soil moisture status in China with the TVDI from March to May 2000 and found a significant negative linear correlation between the TVDI and measured soil moisture from NOAA polar-orbiting satellite data To assess drought in Shandong province in China Gao et al. (2011) integrated TVDI and regional water index (RWI) with Landsat TM / ETM + satellite imagery. Besides, Tao et al. (2011) applied GIS to monitor drought on Tongj in the land of Dafang district in Bijie prefecture of west Guizhou province. Son et al. (2012) illustrated the use of monthly MODIS NDVI and LST data to monitor agricultural drought along with Trop- ical Rainfall Measuring Mission (TRMM) data. This article mainly studies drought monitor- ing in Cu Chi district based on TVDI using LANDSAT infrared thermal imaging material with a spatial resolution (30m -120m) to provide clearer information on changes in surface mois- ture content. In comparison with MODIS and NOAA/AVHRR images, it can be used effec- tively in researching and monitoring drought at the provincial level. The analysis results con- tribute to improving the method of identifying drought risk zoning to help local governments have an overview of droughts and make appro- priate policies and planning of natural resources, contributing to mitigation. local disasters. Be- sides, the results can be used as useful references for research topics related to drought. 2. Materials and Methods 2.1. Study area The study’s objective is to assess the drought situation in Cu Chi district, Ho Chi Minh City, Viet Nam (Fig. 1). Cu Chi is a suburban district located to the northwest of Ho Chi Minh City, situated at the latitude of 10o53’00” to 11o10’00” N and 106o 22’00” to 106o40’00” E. Cu Chi Dis- trict cover an area of 43,496 ha, with a natural area equaling to 20.74% of the city's area. The area has a typical monsoon tropical climate with two seasons: a dry season from November to April with low humidity and high evapotranspi- ration, and a rainy season from May to October with high humidity and low evapotranspiration (ADP, 2010). Tran Thi Thanh Dung et al./Vietnam Journal of Hydrometeorology, 2020 (04): 41-52 43 Applying TVDI based on remote sensing data to evaluate the drought in Cu Chi District                                                                                                                                                                                                                                                                                                                                                                                                     Fig. 1. Map of the pilot study area, Cu Chi District, Ho Chi Minh City, Central Viet Nam 2.2. Data Landsat images (path 124/ row 052) were downloaded from the USGS data server (earth- explorer.usgs.gov) and used in this study. The first and second images were Landsat 5 The- matic Mapper (TM) acquired on 02/13/2005 and 02/11/2010, respectively, while the third and fourth imagery were Landsat 8 (OLI/ TIRS) ac- quired on 01/24/2015 and 02/23/2020. Based on the study objectives, Landsat images were ac- quired during the dry season in Cu Chi district to best show land features, particularly, vegeta- tion and soil moisture those concerning the oc- currence of drought and to avoid overshadowing by too much vegetation (Ayad et al., 2020). 2.3. Methodology In the method section, the research shows the processing of the Landsat data to estimate tem- poral trends of TDVI changes. Firstly, the Land- sat datasets are pre-processed. The TVDI index was then calculated based on NDVI and LST. Satellite Image Processing To calculate the land surface temperature, the first step of the proposed work is to convert the DN (Digital Number) values of band Thermal infrared to at-sensor spectral radiance (Wm-2 m- 1). Landsat 5 TM images can be converted to Top of Atmosphere (TOA) radiances using the following expression (1) (NASA, 2001): where Lmax is the maximum radiance (Wm-2sr- 1mm-1); Lmin is the minimum radiance (Wm-2sr- 1mm-1); Qcal is the DN value of pixel; Qcalmax is the maximum DN value of pixels; Qcalmin is the minimum DN value of pixels. To estimate the LST from the Landsat-8 ther- mal infrared band data, DN of sensors were con- verted to spectral radiance using the following equation (2) (USGS, 2015): where Lλ is Spectral radiance (Watts/(m2* srad*μm)); ML is Radiance increasing scaling issue for the band (RADIANCE _MULT _BAND_n from the metadata); AL is that the Ra- diance additive scaling issue for the band (RA- DIANCE_ADD_BAND_n from the metadata); Qcal is Level one component worth in DN. The next step is to convert the spectral radiance to TOA brightness temperature under the assumption of uniform emissivity by the fol-                                                                                                                                                                                                                      /O  ௅௠௔௫ି௅௠௜ொ௖௔௟௠௔௫ିொ௖௔௟௠ ݈ܳܿܽ െ ݈ܳܿܽ݉݅݊ܮ݉݅݊                                                                                                                                                                                                                                                                  (1)                                                                                                                                                                                                                                                                                           /O  0/4FDO$/                                                                                                                                                                                              (2) 44 Tran Thi Thanh Dung et al./Vietnam Journal of Hydrometeorology, 2020 (04): 41-52 lowing equation (3) (USGS, 2015; Orhan et al., 2014): where TB is Top of Atmosphere Brightness Temperature; Lλ is Spectral radiance (Watts/(m2 *sr*μm)); K1 is Thermal conversion constant for the band (K1_CONSTANT_BAND_nfrom the metadata); K2 is Thermal conversion constant for the band (K2_CONSTANT_BAND_n from the metadata). For obtaining the results in degrees Celsius, the radiation temperature is adjusted by minus 273.15∘C (Xu et al., 2004; Orhan et al., 2014; Avdan and Jovanovska, 2016). Calculation of Land Surface Temperature (LST or Ts) The Top of Atmosphere Brightness Temper- ature was converted to land surface temperature using the following equation (4) (Yuan et al., 2007; Rulinda et al., 2010): where λ is the central band wavelength of emitted radiance; ρ = h*c/σ (1.438*10-2 m*K); σ is the Boltzmann constant (1.38*10-23 J/K); h is the Planck's constant (6.626*10-34 J*s); c is the light velocity (2.998*108 m/s); ε is the sur- face emissivity. Accurate determination of surface tempera- ture is restricted by associate degree correct data of surface emission. The emissivity of the sur- face is controlled by factors like water content, chemical composition, structure, and roughness (Snyder et al., 1998). It will be determined that the contribution of the assorted parts belongs to the pixels in their proportions. The link between LST and NDVI takes into consideration that veg- etation and soils area unit the most surface pro- tect the terrestrial element. The determination of the bottom emissivity is calculated not ab- solutely as prompt by Valor and Caselles (1996): where εv is vegetation emissivity and εs is soil emissivity. For the territory of Vietnam, several studies in Ho Chi Minh City have determined the εv and εs for LANDSAT images corresponding to 0.904 and 0.991 (Van et al., 2009). Pv is the Proportion of Vegetation in a pixel. Pv is calculated according to Carlson and Ripley (1997) by the following equation (6) (Sobrino et al., 2004): Calculation of Normalized Difference Vege- tation Index The “Normalized Difference Vegetation Index” (NDVI) was introduced by Tucker (1979) which is the most prominent vegetation index derived from remote-sensing (satellite) data used to identify and monitor vegetation. The value NDVI ranges between -1 to 1 with posi- tive values for vegetation and negative values for non-vegetative areas. The NDVI is calculated by the following equation (7) (Myneni et al., 1995). where is the reflectance in Near-Infrared band; is the reflectance in Red band. Calculation of Temperature Vegetation Dry- ness Index The triangle method is based on an interpre- tation of the pixel distribution in the LST/NDVI feature space). Land surface temperature is af- fected by many factors such as surface thermal properties, net radiation, evapotranspiration, and vegetation coverage, hence there is no direct re- lationship between LST and soil water content. However, soil moisture is an important factor controlling vegetation canopy temperature and under certain vegetation coverage, soil moisture can indirectly affect canopy temperature. The Ts/NDVI feature space (Fig. 2) is used to illus- trate the relationship between LST, soil mois- ture, and vegetation coverage. In the study of Price (1990) and Carlson et al. (1994), a scatter plot of remotely sensed surface temperature and                                                                                                                                                                                                                                                                                                                                                                               7%  ܭʹ ܮ݊ሺͳ൅ܭͳܮߣ ሻ Ǧ                                                                                                          (3)                                                                                                                                                                                                                                                                                                                                                                                                                                                                                           OQ % % 7/67 7O HU    (4)                                                                                                                                         İ İY3YİVí3Y                                                                                                                                                                                                                                                                                                                                                                                                                                                                   (5)                                                                                                                                                                                                                      VRLO Y YHJ VRLO 1'9, 1'9,3 1'9, 1'9, § · ¨ ¸¨ ¸© ¹                                                                                                                                                                                                                                                                                                                                                                                            (6)